intrinsic side-channel analysis resistance and e cient masking · optimizing masking costs {...

IntroductionSCA metrics

Optimizing masking costs – nonlinear operationsOptimizing masking costs – comparing CAESAR candidates

ConclusionsRadboud University

Intrinsic Side-Channel Analysis Resistance andEfficient Masking

A case study of the use of SCA-related metrics and of designstrategies leading to low-cost masking for CAESAR candidates

Ko Stoffelen

Master thesis presentationAugust 27, 2015

Ko Stoffelen Master thesis presentation 1 / 36




Acknowledgements

• Supervisor: Lejla Batina

• And Kostas Papagiannopoulos

• Second reader: Joan Daemen





Outline

Introduction

SCA metrics

Optimizing masking costs – nonlinear operations

Optimizing masking costs – comparing CAESAR candidates

Conclusions





Side-Channel Analysis





Masking

• Countermeasure against SCA

• Arithmetic vs. Boolean

• Costs quadratic for nonlinear gates, e.g.:

z = x ∧ y → (x ′ = x ⊕ xm)

(y ′ = y ⊕ ym)

z ′ = x ′ ∧ y ′

zm = (xm ∧ y ′)⊕ (ym ∧ x ′)⊕ (xm ∧ ym)





Goals

• How can known metrics be used at the design stage to assessthe intrinsic resistance of ciphers to implementation- anddevice-dependent attacks?

• How can the costs of applying masking countermeasures tociphers be reduced?





Context – CAESAR competition

ACORN ++AE AEGIS AES-CMCC AES-COBRAAES-COPA AES-CPFB AES-JAMBU AES-OTR AEZArtemia Ascon AVALANCHE Calico CBACBEAM CLOC Deoxys ELmD EnchiladaFASER HKC HS1-SIV ICEPOLE iFeed[AES]Joltik Julius Ketje Keyak KIASULAC Marble McMambo Minalpher MORUSNORX OCB OMD PAEQ PAESPANDA π-Cipher POET POLAWIS PRIMATEsPrøst Raviyoyla Sablier SCREAM SHELLSILC Silver STRIBOB Tiaoxin TriviA-ckWheesht YAES






(S-boxes of)

8x8 5x5 4x4

AES Ascon JoltikAES−1 ICEPOLE Joltik−1

iSCREAM Ketje/Keyak LACSCREAM PRIMATE MinalpherSCREAM−1 PRIMATE−1 Prøst

RECTANGLERECTANGLE−1





Traditional S-box design criteria

S-box Width Nonlinearity Degree δ

AES 8 112 7 4iSCREAM 8 96 6 16SCREAM 8 96 5/6 16

Ascon 5 8 2 8ICEPOLE 5 8 4 8Ketje/Keyak 5 8 2 8PRIMATE 5 12 2/3 2

Joltik 4 4 3 4LAC 4 4 3 4Minalpher 4 4 3 4Prøst 4 4 3 4RECTANGLE 4 4 3 4





SCA metrics





Why additional SCA-related criteria?

• SCA highly effective

• Countermeasures only applied to implementations

• Countermeasures expensive (area, speed)

• Perfect countermeasure does not exist

• A lot to gain with an intrinsically more resistant S-box





Existing metrics

Number of measurements Guessing entropySignal-to-noise ratio Confusion coefficientTransparency order Modified transparency orderSuccess rate Second minimum distanceNew signal-to-noise ratio





But. . .

• Metrics take different approaches

• Metrics work under different assumptions (power model,Gaussian noise, . . . )

• Some only applicable in certain cases

• Not all meaningful in design stage





Confusion coefficient

• Intuitively: probability that power analysis attack succeeds

• Result is frequency distribution

• Lower mean ⇒ higher resistance

• Mean only depends on size of S-box

• Higher variance ⇒ higher resistance





Confusion coefficient – first-order

AES−1

AES

iSCREA

M

SCREA

M

Ket

je/K

eyak

SCREA

M−1

ICEPO

LE

Jolti

k

PRIM

ATE−1

PRIM

ATE

LAC

Jolti

k−1

Prøst

Asc

on

REC

TANGLE

REC

TANGLE−1

Min

alph

er0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

cv

HWHDValueWeightedPairs





Confusion coefficient – second-order

AES−1

AES

iSCREA

M

SCREA

M

Ket

je/K

eyak

SCREA

M−1

ICEPO

LE

Jolti

k

PRIM

ATE−1

PRIM

ATE

LAC

Jolti

k−1

Prøst

Asc

on

REC

TANGLE

REC

TANGLE−1

Min

alph

er0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

cv

HWHDValueWeightedPairs





Confusion coefficient conclusions

• Confusion coefficient can deal with low-entropy maskingschemes

• The ranking of the S-boxes according to the confusioncoefficient is mostly preserved by various power consumptionmodels

• The ranking of the S-boxes according to the confusioncoefficient is mostly preserved by higher-order attacks

• Assumption: mean and variance are of similar importance





SCA metrics comparison

Jolti

kLA

C

Jolti

k−1

Prøst

REC

TANGLE

REC

TANGLE−1

Min

alph

er0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

SNRTransparency order

Modified transparency orderSecond minimum distance

Confusion coefficient variance






Ket

je/K

eyak

ICEPO

LE

PRIM

ATE−1

PRIM

ATE

Asc

on2

2.5

3

3.5

4

4.5

5

5.5

6

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7









AES−1

AES

iSCREA

M

SCREA

M

SCREA

M−1

5

6

7

8

9

10

11

12

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4








SCA metrics verdict

• SNR, modified transparency order, and confusion coefficient areconsistent in their predictions

• Second minimum distance a bit less, requires further research

• Metrics behave as they should under various circumstances

• Minalpher, Ascon, SCREAM−1 are suggested to have the mostDPA-resistant S-boxes

• However. . .





SCA metrics verdict

• SCA simulation results do not agree• Usefulness of metrics still unclear

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 9595 1000.9

1

1.1

1.2

1.3

1.4

1.5

1.6

N

r′

AES−1

SCREAM−1

Ascon

Ketje/KeyakJoltikMinalpher





Optimizing masking costs

Nonlinear operations





Multiplicative complexity (MC)

• Recall that the cost of masking nonlinear operations isquadratic in the number of inputs

• Most nonlinear operations in the nonlinear part of the primitive:the S-box

• MC: minimal number of AND/OR gates required to implementfunction

• Goal is to compute the MC of CAESAR S-boxes





Minimizing AND/OR gates

• Existing logic synthesis tools not very helpful• E.g. Espresso, SIS, misII, Logic Friday, ABC, . . .

• Instead: convert to SAT and let SAT solvers do the work

• Converting problem to SAT nontrivial, but feasible





Reducing decisional MC to SAT

q0 = a0 + a1 · x0 + a2 · x1 + a3 · x2 + a4 · x3

q1 = a5 + a6 · x0 + a7 · x1 + a8 · x2 + a9 · x3

t0 = q0 · q1

q2 = a10 + a11 · x0 + a12 · x1 + a13 · x2 + a14 · x3 + a15 · t0

q3 = a16 + a17 · x0 + a18 · x1 + a19 · x2 + a20 · x3 + a21 · t0

t1 = q2 · q3

q4 = a22 + a23 · x0 + a24 · x1 + a25 · x2 + a26 · x3 + a27 · t0 + a28 · t1

q5 = a29 + a30 · x0 + a31 · x1 + a32 · x2 + a33 · x3 + a34 · t0 + a35 · t1

t2 = q4 · q5

y0 = a36x0 + a37 · x1 + a38 · x2 + a39 · x3 + a40 · t0 + a41 · t1 + a42 · t2








My work

• Wrote scripts to generate logic formulas in ANF from S-boxand given MC

• Use tool to convert ANF to CNF

• Let MiniSAT and CryptoMiniSAT do the work on DS clusternode

• Wrote scripts to convert back to S-box implementation





Results

S-box MC S-box MC

AES ≤ 32 PRIMATE−1 ∈ {6, 7, 8, 9, 10}*AES−1 ≤ 32 Joltik 4iSCREAM ≤ 12 Joltik−1 4*SCREAM ≤ 11 LAC 4*SCREAM−1 ≤ 11 Minalpher 5*Ascon 5 Prøst 4ICEPOLE 6* RECTANGLE 4Ketje/Keyak 5 RECTANGLE−1 4*PRIMATE ∈ {6, 7}*





Optimizing masking costs

Comparing CAESAR candidates





High-level operations

• Table lookups

• Bitwise Boolean functions

• Shifts and rotates

• Modular addition/multiplication

• Modular polynomial multiplication





Results

OperationTablelookups

BitwiseBoolean

Shifts/rotates

Mod. add.and mult.

Mod. poly.mult.

AES 256 bytes XOR Fixed XiSCREAM 512 bytes AND,OR,XOR Fixed × mod 256SCREAM 512 bytes AND,OR,XOR × mod 256

Ascon AND,OR,XOR FixedICEPOLE 96 bytes AND,XOR FixedKetje/Keyak AND,XOR FixedPRIMATE 25 bytes XOR Fixed XJoltik 64 bytes XOR Fixed + mod 16 XLAC 16 bytes XOR FixedMinalpher 16 bytes XOR

Prøst AND,XOR FixedRECTANGLE AND,OR,XOR Fixed





Results

• Expected masking costs not so high on average

• Ascon, Ketje, Keyak, LAC, Minalpher, Prøst, and RECTANGLEstand out

• Designers should use operations that are cheap to mask using aBoolean scheme





Conclusions

• SNR, modified transparency order, and confusion coefficientcredible in theory

• However, SCA simulations do not reflect the expectationssuggested by metrics

• For 4- and 5-bit S-boxes, we can find an implementation with aprovably minimum number of AND/OR operations

• Ascon, Ketje, Keyak, LAC, Minalpher, Prøst, and RECTANGLEare expected to have the lowest masking costs





Questions

?


intrinsic side-channel analysis resistance and e cient masking · optimizing masking costs {...

Documents